Skip to content
Download for Mac

Interface Tour

QARK’s interface has six regions: icon sidebar, content sidebar, tab bar, main panel, info panel, and the universal overlay.

Annotated full-window view of QARK with callouts for icon sidebar, content sidebar, tab bar, main panel, info panel Detailed Annotations

The narrow vertical strip on the far left. Each icon switches the content sidebar to a different view.

Top section (views):

IconViewShortcut
ConversationsConversation list, groups, search
ArchiveArchived conversationsCtrl+Backspace / Cmd+Backspace
TrashDeleted conversations with restore
PromptsPrompt library with categories and favoritesCtrl+Shift+P / Cmd+Shift+P
FlowsFlow library with categories and output action badgesCtrl+G / Cmd+G
AgentsAgents and sub-agents (agent-tools)Ctrl+Shift+A / Cmd+Shift+A

Bottom section (actions):

IconAction
Toggle sidebarCollapse or expand the content sidebar (Ctrl+B / Cmd+B)
ThemeCycle between light, dark, and system theme
SettingsOpen the settings dialog

Clicking any view icon auto-expands the sidebar if it’s collapsed.

The panel next to the icon strip. Its contents change based on which icon is active.

The default view. Shows all conversations organized into groups.

Three system groups are always present:

GroupPurpose
GeneralDefault destination for new conversations
SparksConversations created from the overlay’s Spark tab
FlowsConversations created by Flow executions

Create custom groups by right-clicking the sidebar. Drag conversations between groups to reorganize. Groups support custom colors and icons.

The search field at the top filters by title and message content. Results update as you type.

Hold Ctrl / Cmd and click to multi-select conversations. Batch actions: group, archive, delete, clear selection.

Browse and manage reusable prompt templates. Prompts are organized by category with a favorites section at the top. Search across title, description, content, and tags. Import/export prompts as JSON.

Browse and manage execution pipelines. Flows are organized by category with favorites pinned at the top. Each flow card shows its output action (Copy, Paste, Insert Before, Insert After, Open in Chat) and preview mode. Import/export as JSON.

Create and manage agents and sub-agents (agent-tools). Two sections: Agents (full agents with system prompt, model, and tools) and Sub-Agents (agents that other agents can invoke as tools, with typed input/output schemas). Built-in agents are marked with a badge. Import/export as JSON.

Archive holds conversations you want to keep but hide from the main list. Unarchive to restore.

Trash holds deleted conversations. Restore or permanently delete. Empty Trash clears everything.

Every open conversation gets a tab. The tab bar sits above the main panel.

  • Pin — right-click a tab → Pin. Pinned tabs collapse to icon width and persist across restarts.
  • Split view — press Ctrl+\ / Cmd+\ to split the panel. Compare outputs from different models side by side, or reference one conversation while working in another.
  • Overflow — when tabs exceed the bar width, an overflow dropdown appears on the right.
  • Provider colors — each tab shows a subtle top-border accent color from its model’s provider (e.g. violet for Anthropic, green for OpenAI), so you can identify the provider at a glance.

The center of the interface. Contains the message thread and the chat input at the bottom.

Above the input field. Shows the current provider and model (e.g. “OpenRouter / OpenAI: GPT-5.4”). Click to open the model picker — a two-panel modal with provider list on the left and model details (context window, pricing, capabilities) on the right.

Two toggles sit next to the model name:

  • Thinking — enable or disable extended chain-of-thought reasoning. Only appears when the selected model supports it. Blue when active.
  • Thinking visibility — show or hide thinking blocks in the message thread. Useful if you want thinking enabled but don’t need to read the reasoning steps.

The text area where you type prompts. Auto-resizes as you type. Supports pasting images, drag-and-drop files and folders, and @ mentions to attach tools.

Below the input, a toolbar with four buttons:

ButtonWhat it does
Context strategyChoose how conversation history is sent to the model — Auto Compact (default), Last N, First N, All, None, or Token Budget. Highlighted when set to a non-default strategy.
ToolsOpen the tool picker to enable or disable tools for this conversation. Shows built-in tools, MCP server tools, and agent-tools. Highlighted when tools are enabled.
AttachAdd images (for vision), files or folders (for RAG), grab the current Finder/Explorer selection, or pick from clipboard history. Highlighted when files are pending.
PromptsBrowse the prompt library. Search, pick a prompt, fill in any placeholder variables, and insert it into the input field.

The send button (arrow icon) sits on the right. While streaming, it turns into a stop button (square icon) to cancel generation.

A collapsible right-hand panel. Toggle with Ctrl+I / Cmd+I. Three tabs:

TabContents
ConfigAgent, model, system prompt, temperature, max tokens, enabled tools, context strategy — edit any field to override defaults for this conversation only
ArtifactsFiles attached to the conversation — PDFs, images, code, DOCX — with filename, size, and processing status for RAG-indexed documents
InfoConversation metadata: creation date, total messages, cumulative tokens, cumulative cost, models used
Universal overlay — Spark tab with captured text, and Flows tab showing flow categories Detailed Overlay Annotations

A floating panel summoned from any application with a global hotkey. Default: Ctrl+Alt+Space (Windows/Linux) or Cmd+Option+Space (macOS). Configurable in Settings → Sparks & Flows.

The overlay has two tabs — Spark and Flows — for two independent features that share the same overlay surface.

For quick, direct prompts. Highlight text in any app, press the hotkey, and the selected text appears in a collapsible Captured text block at the top (with character count and a dismiss button). Type a prompt in the input field — or use the captured text as-is — and press Enter. The response streams directly into the overlay.

The Spark input has its own model bar (provider / model picker + thinking toggle) and the same toolbar as the main chat input — tools, attachments, and prompt picker. An output action dropdown on the right lets you choose what happens with the result: Copy to Clipboard (default), Paste at Cursor, Insert Before, Insert After, or Open in Chat.

For running pre-configured execution pipelines against the captured text. Switch to the Flows tab and you get a searchable list of all your flows, grouped by category (Analysis, Writing, Code, etc.). Each flow shows its name, description, and output action badge.

Select a flow and it runs immediately with the captured text as input. The input field placeholder changes to “Add input for the flow…” — you can add extra context before sending.

  • Open in main chat button (external link icon) — transfers the conversation to the main QARK window. If streaming is in progress, it continues seamlessly in the main window.
  • Pin button (top right) — keep the overlay open when you click away instead of dismissing on blur.
  • Close button (X) — dismiss the overlay.
  • Press Esc to dismiss.

Spark conversations are saved to the Sparks system group, Flow conversations to the Flows system group. Both are full QARK conversations — open them from the sidebar to continue, attach files, or switch models.